mehulashah

I am gobsmacked by this letter.

Every one of the mathematicians on this letter was likely aware of the progress in AI, and knew this was coming if they were honest with themselves. Some have even helped pushed the agenda forward. Inevitably, there would be a day when AI would find solutions to problems that defined the field. That day has come perhaps sooner than anticipated.

This letter smacks of fear. Their way of learning, work, achievement, and community is threatened by the ever steady pace of technology.

They say:

“But solving problems is only a tool and proxy for achieving the primary goal of conceptual understanding and insight. Forgetting this in the world of AI may turn the tool against the primary goal. Indeed, the mass production at faster and faster pace of "true/false" statements could destroy fertile ground instead of breathing life into new ideas.”

Indeed, but the mathematics community needs to look inward not outward.

Achievement has long been defined by solving problems that other mathematicians and scientists have found crucial for understanding shapes, numbers, and natural phenomena.

The race to solve these hasn’t always been collaborative, cordial, or healthy. Too often, individuals have worked in secret or in small groups in order to claim priority and the accolades that come along. Further, I’ve even seen worse behavior not worth getting into here.

Now AI can solve these problems, oftentimes without building the abstractions that we need for understanding. And it is increasingly easier to collaborate in large groups without needing to trust everyone’s ability because we can check the truth of their work and incorporate it into the larger effort.

Certainly, the mathematical community and AI are misaligned. But, it is unclear that AI needs to change. Rather, I argue, it is the mathematical community that will need to change. Instead of focusing on the teleological pursuit of the truth of blessed problems, we need to reward the concise explanations and beautiful theories that give clear understanding. In turn, we can use AI to help us collaborate better and better our understanding.

For all the naysayers in this thread, I gotta say you’re wrong. Vint is a class act. Humble, helpful, and optimistic. Not to mention one of the most impactful computer scientists of our generation.

The first and seminal Indian tech blogger. I looked up to him. RIP.

He makes it sound simple, and in some ways, it is. You have to not only build the thing, but also find the thing that people want. It’s the latter that’s as hard if not harder than the former.

In addition, the converse is not true. Just because you’ve found something that grows fast and in large market, doesn’t mean you’ll become a billionaire. With all humility, I’ve been lucky to have done that twice, but in a large company. I’m not complaining, I’m just saying that doesn’t necessarily make you a billionaire.

To be clear: yes, it’s possible, it’s not too late. Read the book.

Eric - I've worked for NASA, ATT, IBM, HP, Amazon, and Google, not to mention a couple of startups that I started in between. None of them (except the startups, but they were brief) stayed true to their original mission. I haven't read your new book, but IMO, it's because the founders leave and the next leadership don't share the vision or values of the founders in the same way. After all, a company is a collaboration among people who want to make a contribution. When the people change, the company changes. It's inevitable.

That said, you seem to have archetypes above Costco, Patagonia, and Novo Nordisk that avoided it.

Can you comment on not what it takes to build such a company, but rather how to transform companies like those that I worked for into ones that resist gravity? Or is it too late?

Why not?

This administration has made it clear in no uncertain words or actions. They don’t want immigrants. And if you think this is bad policy, please stop voting for them. Please vote for the alternative candidates. This is the easiest way to fix this nonsense.

I just want to state for the record that this will fail harder than Google Plus.

This is cool.

This is a great article and I share its goals. But, it ignores something fundamental about humans as a collective — capitalism. Capitalism is what got us here and is at odds with first understanding and then building. We’ve done this before with other technologies because that’s how our societies have learned to grow and collaborate at large scale. First build and build to its limits. Then understand and fix if necessary. Nothing new here, but stopping the trend toward epistemic collapse requires building incentives into the system for us humans to coevolve with AI.

This is sick.

So, I’m left wondering why he did it the hard way.

I learned the first time I got slashdotted:

1. Don’t engage in public with an antagonistic or upset user or reviewer.

2. The thread will unroll itself, and the immaterial ones will die out on their own.

I spent 9 months at IBM in 1999. At that time, Lou’s legacy had already been solidified. He saved IBM. While not everyone agreed with his decisions, there was a culture of both honesty to customers and innovation that permeated the company. In contrast, look what happened to HP without such great leadership. Once a shining light in Silicon Valley, it turned into a shell of its former self.

He right. I’ve seen poverty in India, but I misunderstood it.

There was a 12 year old kid who guided our boat down the Narmada after we spread my Dads ashes. He was not in school because he wanted money.

I told him I’d pay him double and continue to pay him for his days work, if he’d go back to school during the day and only row boats at night.

He said no. Just give me what you owe me.

He had no hope that education in the government schools would meaningfully change anything for him. Poverty is not a single static state. It’s a negative feedback loop that requires systemic change to get out of.

Certainly research has made it into product with the help of the innovators that created the research. The dial is turned further here where the research ideas have yet to be tried and vetted. The research begins in the startup. Even in the dotcom era, the research prototypes were vetted in the conferences and journals before taking the risk to build production systems. This is no longer the case. The experiments have yet to be run.

Most of the folks on this topic are focused on Meta and Yann’s departure. But, I’m seeing something different.

This is the weirdest technology market that I’ve seen. Researchers are getting rewarded with VC money to try what remains a science experiment. That used to be a bad word and now that gets rewarded with billions of dollars in valuation.

This site is a stroll down memory lane for me. I had one of the calculator watches as a kid. Used it everywhere, especially when shopping or going out to eat with parents. It was cool to cross check totals and check tax calculations. I was a nerdy kid and craved math.

This smacks so much of a Silicon Valley episode. “Pre-idea individuals” … Sounds like they want people with no opinions. Next we will say stuff like “No thought personas”

I like this post. I see this confusion all the time! What’s the difference between ChatGPT and gpt-5 or gpt-4o, and so on. OpenAI’s carefully crafted naming schemes don’t help. Though, I come from AWS so glass houses.

Anyway, agents are control systems that using planning, tools, and a collection of underlying models. ChatGPT is an agent. What kind? The kind optimized for the general user looking to do work with public knowledge. That’s the best definition I can come up with.

Anyway, let’s make sure people understand the difference between AI systems and AI models. The former is where a lot of startup activity will be for a decade. The latter will be in the hands of a few well funded behemoths.

Oh dear. Should I comment on this?

I’m surprised by the argument. It’s not wrong. You need more data, but that presumes that the task is to pre-train on data. Additional compute is also useful for unearthing tacit capabilities in the models. This requires inference time scaling and post training usually on specific downstream tasks using RL. Sure that generates data, but it’s not the same as the Internet, and can be scaled.

I took 6.034 from Winston and still have the lecture notes and book. Though dated, they remind me of what was great about MIT. The constant change, upheaval, search for scientific truth, and desire to help humankind. RIP Patrick Winston.

Trust me. I’m usually the last to jump on a bandwagon. That said, this is not just my take, but the take of many others that I trust. Andrej Karpathy, Joseph Hellerstein, etc.

Agree. The non-determinism is a problem. Perhaps systems thinking can help here. After all, traditional computer systems often behave non-deterministically and we’ve learned to make them more reliable.

I shared your opinion for a while. But, that’s not what’s happening. People are using them for everything. When they do, expectations are set. Vendors will adjust and so will the rest of the industry. It’s happening.

I do believe it’s time for systems folks to take a hard look at building systems abstractions on top of LLMs as they did 50 years ago on top of CPUs. LLMs are the new universal computing machines, but we build with them like we built computers in the 1950s - one computer at a time.

100%. There’s no point in nitpicking on this post. There’s an outsized return on exercise and it’s measurable. People don’t get — especially young people — that exercise is like eating, sleeping, and pooping. Your body needs it in regular intervals otherwise its carefully balanced system goes out of whack.

So, what’s the right answer and how do you know? The only way to know is to go to some primary source.